It is possible that this skip path causes TCP to end up into an
invalid state where ca_state was left to CA_Open while some
segments already came into sacked_out. If next valid ACK doesn't
contain new SACK information TCP fails to enter into
tcp_fastretrans_alert(). Thus at least high_seq is set
incorrectly to a too high seqno because some new data segments
could be sent in between (and also, limited transmit is not
being correctly invoked there). Reordering in both directions
can easily cause this situation to occur.
I guess we would want to use tcp_moderate_cwnd(tp) there as well
as it may be possible to use this to trigger oversized burst to
network by sending an old ACK with huge amount of SACK info, but
I'm a bit unsure about its effects (mainly to FlightSize), so to
be on the safe side I just currently fixed it minimally to keep
TCP's state consistent (obviously, such nasty ACKs have been
possible this far). Though it seems that FlightSize is already
underestimated by some amount, so probably on the long term we
might want to trigger recovery there too, if appropriate, to make
FlightSize calculation to resemble reality at the time when the
losses where discovered (but such change scares me too much now
and requires some more thinking anyway how to do that as it
likely involves some code shuffling).
This bug was found by Brian Vowell while running my TCP debug
patch to find cause of another TCP issue (fackets_out
miscount).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
[ 63.531438] =================================
[ 63.531520] [ INFO: inconsistent lock state ]
[ 63.531520] 2.6.26-rc4 #7
[ 63.531520] ---------------------------------
[ 63.531520] inconsistent {softirq-on-W} -> {in-softirq-W} usage.
[ 63.531520] tcpsic6/3864 [HC0[0]:SC1[1]:HE1:SE0] takes:
[ 63.531520] (&q->lock#2){-+..}, at: [<c07175b0>] ipv6_frag_rcv+0xd0/0xbd0
[ 63.531520] {softirq-on-W} state was registered at:
[ 63.531520] [<c0143bba>] __lock_acquire+0x3aa/0x1080
[ 63.531520] [<c0144906>] lock_acquire+0x76/0xa0
[ 63.531520] [<c07a8f0b>] _spin_lock+0x2b/0x40
[ 63.531520] [<c0727636>] nf_ct_frag6_gather+0x3f6/0x910
...
According to this and another similar lockdep report inet_fragment
locks are taken from nf_ct_frag6_gather() with softirqs enabled, but
these locks are mainly used in softirq context, so disabling BHs is
necessary.
Reported-and-tested-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
In xt_connlimit match module, the counter of an IP is decreased when
the TCP packet is go through the chain with ip_conntrack state TW.
Well, it's very natural that the server and client close the socket
with FIN packet. But when the client/server close the socket with RST
packet(using so_linger), the counter for this connection still exsit.
The following patch can fix it which is based on linux-2.6.25.4
Signed-off-by: Dong Wei <dwei.zh@gmail.com>
Acked-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The field was supposed to allow the creation of an anycast route by
assigning an anycast address to an address prefix. It was never
implemented so this field is unused and serves no purpose. Remove it.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make nlmsg_trim(), nlmsg_cancel(), genlmsg_cancel(), and
nla_nest_cancel() void functions.
Return -EMSGSIZE instead of -1 if the provided message buffer is not
big enough.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Also removes an unused policy entry for an attribute which is
only used in kernel->user direction.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Also removes an obsolete check for the unused flag RTCF_MASQ.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
No need to compute copy twice in the frags loop in
dma_skb_copy_datagram_iovec().
Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The neighbor table time of last use information is returned in the
incorrect unit. Kernel to user space ABI's need to use USER_HZ (or
milliseconds), otherwise the application has to try and discover the
real system HZ value which is problematic. Linux has standardized on
keeping USER_HZ consistent (100hz) even when kernel is running
internally at some other value.
This change is small, but it breaks the ABI for older version of
iproute2 utilities. But these utilities are already broken since they
are looking at the psched_hz values which are completely different. So
let's just go ahead and fix both kernel and user space. Older
utilities will just print wrong values.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bad type/protocol specified result in sk leak.
Fix is simple - release the sk if bad values are given,
but to make it possible just to call sk_free(), I move
some sk initialization a bit lower.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Jarek Poplawski <jarkao2@gmail.com>
There is only one function in AX25 calling skb_append(), and it really
looks suspicious: appends skb after previously enqueued one, but in
the meantime this previous skb could be removed from the queue.
This patch Fixes it the simple way, so this is not fully compatible with
the current method, but testing hasn't shown any problems.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's logic in __rfcomm_dlc_close:
rfcomm_dlc_lock(d);
d->state = BT_CLOSED;
d->state_changed(d, err);
rfcomm_dlc_unlock(d);
In rfcomm_dev_state_change, it's possible that rfcomm_dev_put try to
take the dlc lock, then we will deadlock.
Here fixed it by unlock dlc before rfcomm_dev_get in
rfcomm_dev_state_change.
why not unlock just before rfcomm_dev_put? it's because there's
another problem. rfcomm_dev_get/rfcomm_dev_del will take
rfcomm_dev_lock, but in rfcomm_dev_add the lock order is :
rfcomm_dev_lock --> dlc lock
so I unlock dlc before the taken of rfcomm_dev_lock.
Actually it's a regression caused by commit
1905f6c736 ("bluetooth :
__rfcomm_dlc_close lock fix"), the dlc state_change could be two
callbacks : rfcomm_sk_state_change and rfcomm_dev_state_change. I
missed the rfcomm_sk_state_change that time.
Thanks Arjan van de Ven <arjan@linux.intel.com> for the effort in
commit 4c8411f8c1 ("bluetooth: fix
locking bug in the rfcomm socket cleanup handling") but he missed the
rfcomm_dev_state_change lock issue.
Signed-off-by: Dave Young <hidave.darkstar@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since mmc_spi.h uses irqreturn_t type, it should include appropriate
header, otherwise build will break if users didn't include it (some of
them do not use interrupts).
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney/blackfin-2.6:
8250 Serial Driver: revert extra IRQ flag definition patch
Blackfin arch: update anomaly headers from toolchain trunk
Blackfin arch: Remove bad and usless code
Blackfin arch: Fix bug - set corret SSEL and IRQ to enable AD7877 on BF527
Blackfin arch: Fix typo. it should be _outsw_8
Blackfin arch: Cleanup no functional changes
When we introduced support for KBUILD_EXTRA_SYMBOLS
we started to include the externam module's kbuild
file when doing the final modpost step.
As external modules often do:
ccflags-y := -I$(src)
We had problems because $(src) was unassinged and
gcc then used the next parameter for -I resulting in
strange build failures.
Fix is to assign $(src) and $(obj) when building
external modules.
This fixes: http://bugzilla.kernel.org/show_bug.cgi?id=10798
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Tvrtko <tvrtko.ursulin@sophos.com>
Cc: Andrea Arcangeli <andrea@qumranet.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
As Russell pointed out, original patch will break some serial configurations
because of the dependency of the <asm/serial.h> header file.
Revert it first and try to find out other solution later
Cc: Javier Herrero <jherrero@hvsistemas.es>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Russell King <rmk+lkml@arm.linux.org.uk>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
AD7877 use SSEL_2 (P9.9) and IRQ_PF8 (P9.14) on BF527
- populating JP3 to enable STAMP
- disable SW11.2 to disconnect SSEL_2/PF12 from Rotary NAND
- disable SW18.1 to disconnect SSEL_2 from MAX1233 touchscreen chip
Signed-off-by: Bryan Wu <cooloney@kernel.org>
The node for DMA2 in the MPC8610 HPCD device tree has the wrong compatible
properties. This breaks the DMA driver and the sound driver.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Currently ext4 and fuse fail to link if modular:
ERROR: "copy_page" [fs/fuse/fuse.ko] undefined!
ERROR: "empty_zero_page" [fs/ext4/ext4dev.ko] undefined!
make[3]: *** [__modpost] Error 1
make[2]: *** [modules] Error 2
make[1]: *** [sub-make] Error 2
While arch ppc exists it may as well compile, so this exports those
symbols (which are already exported in arch/powerpc).
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Gcc might re-order MMIO accessors vs. surrounding consistent
memory accesses, which is a "bad thing", and could break drivers.
This fixes it by adding a "memory" clobber to the MMIO accessors,
which should prevent gcc from doing that reordering.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The following patch is seems to fix the tulip suspend/resume panic:
http://bugzilla.kernel.org/show_bug.cgi?id=8952#c46
My attempts at a cleaner patch failed and Pavel thinks this is OK.
Original from: kernelbugs@tap.homeip.net
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
When using 4+ GB RAM and SWIOTLB is active, the driver corrupts
memory by writing an skb after the relevant DMA page has been
unmapped. Although this doesn't happen when *not* using bounce
buffers, clearing the pointer to the DMA page after unmapping
it fixes the problem.
http://marc.info/?t=120861317000005&r=2&w=2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Jay Cliburn <jacliburn@bellsouth.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Because we cache the last failed-to-xmit packet, if there are no
packets queued behind that one we may never send it (reproduced here
as TCP stalls, "cured" by an outgoing ping).
Cc: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
If we fail to transmit a packet, we assume the queue is full and put
the skb into last_xmit_skb. However, if more space frees up before we
xmit it, we loop, and the result can be transmitting the same skb twice.
Fix is simple: set skb to NULL if we've used it in some way, and check
before sending.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Commit 5a0a92e67b mentions len < ETH_ZLEN
is true for ARP packets. This obviously is not unlikely.
Signed-off-by: Cesar Eduardo Barros <cesarb@cesarb.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
After request_dma() succeeding, any error path should do free_dma().
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This is normal when the external link is down so don't report it as an error.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This patch silences the build warnings concerning o2net_init_nst()
and friends when building without CONFIG_DEBUG_FS enabled.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
This patch silences the build warnings concerning dlm_debug_init()
and friends when building without CONFIG_DEBUG_FS enabled.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
This patch silences the build warnings concerning o2net_debugfs_init()
and friends when building without CONFIG_DEBUG_FS enabled.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
The static structure describing the userspace cluster plugin for ocfs2
was named 'user_stack', which is a real pain when people are grep(1)ing
the tree for the program stack object 'user_stack'. Change the name to
something distinct and namespaced.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
lguest: notify on empty
virtio: force callback on empty.
virtio_blk: fix endianess annotations
virtio_config: fix len calculation of config elements
virtio_net: another race with virtio_net and enable_cb
virtio: An entropy device, as suggested by hpa.
virtio_blk: allow read-only disks
lguest: fix ugly <NULL> in /proc/interrupts
virtio: set device index in common code.
virtio: virtio_pci should not set bus_id.
virtio: bus_id for devices should contain 'virtio'
Fix crash in virtio_blk during modprobe ; rmmod ; modprobe
lguest: use ioremap_cache, not ioremap
* Wim Van Sebroeck <wim@iguana.be> wrote:
> Author: Jordan Crouse <jordan.crouse@amd.com>
> Date: Mon Jan 21 10:07:00 2008 -0700
>
> [WATCHDOG] Add a watchdog driver based on the CS5535/CS5536 MFGPT timers
-tip testing found the following build failure on latest -git:
drivers/watchdog/geodewdt.c: In function 'geodewdt_probe':
drivers/watchdog/geodewdt.c:225: error: too many arguments to function 'geode_mfgpt_alloc_timer'
make[1]: *** [drivers/watchdog/geodewdt.o] Error 1
make: *** [drivers/watchdog/geodewdt.o] Error 2
with this config:
http://redhat.com/~mingo/misc/config-Fri_May_30_15_19_52_CEST_2008.bad
find the fix below.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Jordan Crouse <jordan.crouse@amd.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When Greg "fixed" the sysfs usage of that driver a while back, he seem
to have introduced a bug where the quotes are added around the name of
our specific sysfs files, thus breaking the user space tool.
This fixes it. Tested DLPAR operations on a POWER6 machine successfully.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
David S. Miller
Ilpo Järvinen
Jarek Poplawski
Dong Wei
Thomas Graf
Brice Goglin
Stephen Hemminger
Adrian Bunk
Pavel Emelyanov
Dave Young
Linus Torvalds
Anton Vorontsov
Alexey Dobriyan
Sam Ravnborg
Bryan Wu
Mike Frysinger
Michael Hennerich
Timur Tabi
Tony Breeds
Benjamin Herrenschmidt
Grant Grundler
Li Yang
Joakim Tjernlund
Rusty Russell
Brice Goglin
Cesar Eduardo Barros
Wang Chen
Ben Hutchings
Sunil Mushran
Joel Becker
Ingo Molnar